Project
Agentic Product Content Pipeline
A multi-agent pipeline that builds an entire product catalogue from raw supplier data.
The problem
Cataloguing products by hand does not scale. As supplier feeds grow, listings get slower to produce and less consistent, and the backlog only ever gets longer.
What I built
A pipeline of cooperating agents that ingest supplier data, extract and enrich attributes, validate against a schema, and publish, with a Pinecone-backed retrieval layer classifying each product into the right place in the taxonomy. Output is structured and quality-gated at every step.
The result
A small team ships a catalogue that would otherwise need a department, at a consistency manual entry never reaches.
Stack
PythonOpenAIPinecone (RAG)DockerStructured output validation